BU Partners on DOE Genesis Mission award to Advance AI-driven Plasma Physics Research
Chuanfei Dong and Brian Kulis to develop AI tools that accelerate plasma physics research
Boston University is a collaborating institution on a new U.S. Department of Energy (DOE) Genesis Mission award led by Princeton University and announced by the DOE this July. The Genesis Mission supports the development and demonstration of AI-enabled scientific workflows to accelerate breakthroughs in energy, discovery science, and national security.
The multi-institutional project, “Towards a FLARE Digital Twin: Accelerating Experimental Research Using Neural Networks,” will use artificial intelligence (AI) to accelerate plasma physics research. It will establish the foundation for a digital twin—a virtual, AI-powered representation—of the Facility for Laboratory Reconnection Experiments (FLARE) at Princeton Plasma Physics Laboratory (PPPL).
FLARE is an experimental facility that studies magnetic reconnection, the process by which magnetic field lines break apart and reconnect, releasing energy. This fundamental plasma process powers solar flares, drives space weather, and underpins the operation of fusion devices.
Chuanfei Dong, an assistant professor of astronomy and electrical and computer engineering, serves as Boston University’s institutional principal investigator. Brian Kulis, a professor with appointments in engineering, computer science, and computing and data sciences, serves as co-investigator. Both are affiliated faculty with BU’s Hariri Institute for Computing.
The Boston University team will lead one of the project’s central research tasks by developing AI models trained on large-scale kinetic plasma simulations that capture the behavior of individual charged particles. The models will reconstruct global plasma properties, including density and temperature, from sparse measurements collected at only a limited number of locations within the plasma. Because plasma diagnostics are expensive, the AI approach is expected to reconstruct plasma information well beyond what instruments directly measure, reducing the need for costly additional diagnostics while shortening data analysis from months to hours.
The BU team will also contribute to the project’s second research task, led by PPPL, which uses AI to automatically identify and characterize plasmoids in FLARE experimental data. These multi-scale magnetic structures form during magnetic reconnection and provide important clues about how energy is released. The AI approach will enable faster, more consistent analysis of experimental results.
Over the longer term, the researchers aim to develop a universal, AI-enabled foundation model of magnetic reconnection spanning laboratory experiments, fusion devices, space plasmas, the Sun, and other astrophysical environments.
Learn more about this project on the Princeton University website here.